Tuesday, September 6, 2011
Editors note: This post is part of a series that explores the top ten reasons why customers trust Google with their business data. A complete top ten list can be found here.
When users think of Google Apps, they often think of their Gmail inboxes or collaborating on documents in real time with others. They often don’t think of what’s going on behind the scenes. Our cloud computing data centers offer our customers scalability and reliability across all of our products and websites, supporting millions of businesses on Google Apps and over 1 billion Internet searches every day. Our pure and proven cloud offers Apps customers significant data protections that would be hard for those customers to achieve on their own. It’s also the infrastructure that we use to run our own business.
As we’ve grown, we’ve developed an expertise around building data centers and protecting the data stored in them. The machines in the data centers that run our applications are built to our own specifications, including ones focused on security. The hardware is limited to what is necessary for the applications to run, and eliminates unnecessary components such as peripheral connectors or video cards. Similarly, the software that we run on the machines is a specialized, stripped-down version of the Linux operating system leaving out any unnecessary software code such as device drivers. This approach helps provide a computing environment that is less prone to vulnerabilities, compared to typical on-premise, so called “private cloud” or hybrid IT environments.
The services we offer are first and foremost Internet-based applications and platforms. We were born on the Internet, not on a single computer or server. We've published some of our core underlying technologies such as BigTable, the SPDY protocol, Google FIle System (GFS) and MapReduce. The last two of which have gone on to inspire Hadoop, the Apache open source framework that underpins many leading cloud or big data applications. Googlers Luiz André Barroso and Urs Hölzle even wrote a mini-book about some of Google’s approaches, entitled “The Datacenter as a Computer: An Introduction to the Design of Warehouse-Scale Machines”.
Lots of users leads to lots of network traffic that allows us some significant advantages in terms of security. For instance, the spam filtering in Gmail gains rapid visibility into emerging and evolving spam and virus threats, which in turn helps us to block the vast majority of them. This kind of large scale Internet infrastructure also typically provides better protection from denial of service type attacks. It also puts us in a position to spot malicious traffic and help protect users from malware.
Unprecedented global scale would not matter without the ability to reliably deliver business critical services. That is another powerful feature of Google’s technology and process discipline. We’ve built our platform to withstand expected hardware failure, relying on software and highly automated processes in order to support a 99.9% uptime SLA that has no maintenance window. In 2010 Gmail uptime was 99.984% and we are over 99.99% for the first half of 2011. This is an approach you fundamentally can’t take with traditional on premise IT systems.
Running data centers at this kind of scale takes energy, but as a carbon-neutral company we strive to use as little as possible - in fact, our facilities use half the energy of a typical data center. You can read more about our efficiency efforts and our approach to purchasing renewable energy.
In just the 4.5 years I’ve been at Google, I’ve seen quite a few generational changes in the kit we run, be it “simple” things like sheet metal for servers to something more complex like our motherboards, or something even more fluid and complicated like our various software layers. Through all those upgrades, build outs, and migrations, the focus on reliability remains. This is something that keeps me coming back to work day after day, and drives me to help others understand the value we can add to protecting their data and powering their businesses.